Research in deep learning for multi-speaker source separation has received a boost in the last years. However, most studies are restricted to mixtures of a specific number of speakers, called a specific scenario. While some works included experiments for different scenarios, research towards combining data of different scenarios or creating a single model for multiple scenarios have been very rare. In this work it is shown that data of a specific scenario is relevant for solving another scenario. Furthermore, it is concluded that a single model, trained on different scenarios is capable of matching performance of scenario specific models.

A wide variety of neural-network architectures have been proposed for the task of Chinese word segmentation. Surprisingly, we find that a bidirectional LSTM model, when combined with standard deep learning techniques and best practices, can achieve better accuracy on many of the popular datasets as compared to models based on more complex neural-network architectures. Furthermore, our error analysis shows that out-of-vocabulary words remain challenging for neural-network models, and many of the remaining errors are unlikely to be fixed through architecture changes. Instead, more effort should be made on exploring resources for further improvement.

Fuzz testing, or "fuzzing," refers to a widely deployed class of techniques for testing programs by generating a set of inputs for the express purpose of finding bugs and identifying security flaws. Grey-box fuzzing, the most popular fuzzing strategy, combines light program instrumentation with a data driven process to generate new program inputs. In this work, we present a machine learning approach that builds on AFL, the preeminent grey-box fuzzer, by adaptively learning a probability distribution over its mutation operators on a program-specific basis. These operators, which are selected uniformly at random in AFL and mutational fuzzers in general, dictate how new inputs are generated, a core part of the fuzzer's efficacy. Our main contributions are two-fold: First, we show that a sampling distribution over mutation operators estimated from training programs can significantly improve performance of AFL. Second, we introduce a Thompson Sampling, bandit-based optimization approach that fine-tunes the mutator distribution adaptively, during the course of fuzzing an individual program. A set of experiments across complex programs demonstrates that tuning the mutational operator distribution generates sets of inputs that yield significantly higher code coverage and finds more crashes faster and more reliably than both baseline versions of AFL as well as other AFL-based learning approaches.

We propose a benchmark for 6D pose estimation of a rigid object from a single RGB-D input image. The training data consists of a texture-mapped 3D object model or images of the object in known 6D poses. The benchmark comprises of: i) eight datasets in a unified format that cover different practical scenarios, including two new datasets focusing on varying lighting conditions, ii) an evaluation methodology with a pose-error function that deals with pose ambiguities, iii) a comprehensive evaluation of 15 diverse recent methods that captures the status quo of the field, and iv) an online evaluation system that is open for continuous submission of new results. The evaluation shows that methods based on point-pair features currently perform best, outperforming template matching methods, learning-based methods and methods based on 3D local features. The project website is available at bop.felk.cvut.cz.

This work compares user collaboration with conversational personal assistants vs. teams of expert chatbots. Two studies were performed to investigate whether each approach affects accomplishment of tasks and collaboration costs. Participants interacted with two equivalent financial advice chatbot systems, one composed of a single conversational adviser and the other based on a team of four experts chatbots. Results indicated that users had different forms of experiences but were equally able to achieve their goals. Contrary to the expected, there were evidences that in the teamwork situation that users were more able to predict agent behavior better and did not have an overhead to maintain common ground, indicating similar collaboration costs. The results point towards the feasibility of either of the two approaches for user collaboration with conversational agents.

In a dialog, there can be multiple valid next utterances at any point. The present end-to-end neural methods for dialog do not take this into account. They learn with the assumption that at any time there is only one correct next utterance. In this work, we focus on this problem in the goal-oriented dialog setting where there are different paths to reach a goal. We propose a new method, that uses a combination of supervised learning and reinforcement learning approaches to address this issue. We also propose a new and more effective testbed, permuted-bAbI dialog tasks, by introducing multiple valid next utterances to the original-bAbI dialog tasks, which allows evaluation of goal-oriented dialog systems in a more realistic setting. We show that there is a significant drop in performance of existing end-to-end neural methods from 81.5% per-dialog accuracy on original-bAbI dialog tasks to 30.3% on permuted-bAbI dialog tasks. We also show that our proposed method improves the performance and achieves 47.3% per-dialog accuracy on permuted-bAbI dialog tasks.

When I was 15 I discovered The Smiths, a band whose name had by then long been synonymous with misery. But it was Morrissey's unique style of being miserable – coquettish and laced with Northern English humour, flipping between self-pity and irony – that appealed to my teenage self. I'd always cry at the same points in each song: the end of Hand in Glove, the chord changes before the chorus of Girl Afraid, the line in The Queen is Dead where he sings "we can go for a walk where it's quiet and dry". I'm still not sure why the last one had such an effect. Two decades later, Spotify has built an algorithm that aims to quantify the amount of sadness in a music track.

Do you get down to Jackson Five, or is Stravinsky more your style? Artificial intelligence (AI) that predicts taste in music might seem stranger than fiction, but researchers at Jönköping University in Sweden and Maastricht University in the Netherlands believe they've cracked the code. In a paper published on the preprint server Arxiv.org, the team described a system that considers a person's listening behaviors and, using machine learning algorithms and psychological models, infers their "musical sophistication." "Psychological models are increasingly being used to explain … behavioral traces," the team wrote. "The use of domain dependent psychological models allows for more fine-grained identification of behaviors [like music listening] and provide a deeper understanding behind the occurrence of those behaviors."

Artificial intelligence is changing the finance industry, with some early big movers monetizing their investments in back-office AI applications. But as this trend widens, new systemic and security risks may be introduced in the financial system. These are some of the findings of a new World Economic Forum report, The New Physics of Financial Services – How artificial intelligence is transforming the financial ecosystem, prepared in collaboration with Deloitte. "Big financial institutions are taking a page from the AI book of big tech: They develop AI applications and make them available as a'service' through the cloud," said Jesse McWaters, AI in Financial Services Project Lead at the World Economic Forum. "It is turning what were historically cost centres into new source of profitability, and creating a virtuous cycle of self-learning that accelerates their lead." The report, which draws on interviews and workshops with hundreds of financial and technology experts, observes that the "size of the prize" driven through these as-a-service offerings and other applications of AI is much larger than that of the more narrow applications that drive efficiency through the automation of human effort.

KSI and Logan Paul are about to have one of the most anticipated boxing matches in the world. And they'll be doing the entire thing on the internet. The YouTubers have promised "the biggest internet event in history" and it looks set to be watched by a huge number of people. Actually watching it is relatively easy, for a boxing match. It will be streamed on YouTube, as you might expect, and will cost $10 or £7.50.